Machine Learning Investigate Different Time Frames To Root Distinctive Habits

At first, getting dressed in athletic attire and travelling to the gym may seem like a chore. When you finally start going to the gym regularly, you might be eager to drop by your Zumba class or go for a treadmill run. The length of time it takes to develop a gym habit has been revealed by a recent study by social scientists at Caltech: on average, it takes roughly six months.

The same study also examined how long it typically takes for healthcare professionals to develop the habit of washing their hands.

“There is no magic number for habit formation,” asserts Anastasia Buyalskaya (PhD ’21), who is currently an assistant professor of marketing at HEC Paris. Colin Camerer, the Robert Kirby Professor of Behavioural Economics at Caltech, the director and leadership chair of the T&C Chen Centre for Social and Decision Neuroscience, as well as scholars from the Universities of Chicago and Pennsylvania, are additional authors of the study, which was published in the journal Proceedings of the National Academy of Sciences. Xiaomin Li, a former graduate student and postdoctoral researcher at Caltech (MS ’17, PhD ’21), is also an author.

“You may have heard that it takes about 21 days to form a habit, but that estimate was not based on any science,” claims Camerer. “Our work supports the notion that habit formation speed varies depending on the behaviour in question and a variety of other factors,” the authors write.

The study is the first to examine habit formation using machine learning technologies. Tens of thousands of people who were either swiping their badges to enter their gym or washing their hands during hospital shifts provided the researchers with massive data sets to analyse using machine learning. The researchers collaborated with 24 Hour Fitness for the study on the gym, and they collaborated with a business that employed radio frequency identification (RFID) technology to track hospital hand-washing practises for the study on hand-washing. Over 30,000 gym users were tracked over a period of four years, and over 3,000 hospital employees were tracked during nearly 100 shifts.

“With machine learning, we can observe hundreds of context variables that may be predictive of behavioural execution,” says Buyalskaya. “The machine learning does the work for us to find the relevant ones, so you don’t necessarily have to start with a hypothesis about a specific variable,” the author explains.

The researchers were also able to examine people over time in their natural contexts thanks to machine learning; in most earlier studies, participants were only required to complete surveys.

The research revealed that other factors, like the time of day, had little bearing on developing a gym habit. There were other considerations, such as one’s previous actions. For instance, for 76% of gym-goers, how long it had been since their last visit was a significant indicator of whether they would attend again. In other words, a person’s likelihood of developing the habit of going to the gym decreased with the length of time since their last visit. Sixty-nine percent of gym-goers preferred to visit the facility on the same days each week, with Monday and Tuesday drawing the largest crowds.

The researchers examined data from healthcare personnel who were required to wear RFID badges that tracked their hand-washing behaviour for the study’s hand-washing component. We treat the introduction of the RFID technology as a “shock” and think that they may need to rebuild their habits from the moment they use the technology, according to Buyalskaya. It’s likely that some health workers already had the habit before we saw them.

“Overall, we are seeing that machine learning is a powerful tool to study human habits outside of the lab,” says Buyalskaya.

What can machine learning teach us about the establishment of habits? Evidence from Exercise and Hygiene” was financed by the Behaviour Change for Good Initiative, the Tianqiao and Chrissy Chen Institute for Neuroscience, and the Ronald and Maxine Linde Institute of Economics and Management Sciences at Caltech.